Automatización de una unidad de bombeo de 600HP y diseño de un controlador entrenado con sistemas de inferencia difusa basado en redes adaptativas.

Abstract

The present work had as objective the implementation of the automation of a 600HP pumping unit for oil derivatives and the controller design trained by Adaptive neuro fuzzy inference system- ANFIS; the development was focused on two phases; The first one, was the implementation of the automation, where the P & ID was development, the equipment selection by each level of the CIM model, the PLCs and the human machine interface (HMI) programing with Wonderware System Platform, took as reference the cycle of life ANSI / ISA-101.01.2015. Inasmuch as, the ANFIS controller design phase showed the design methodology, initiated with the data capture with Lab View-My Rio, with this data a plant model by neuronal network NoLineal auto-regressive exogenous -NARX and ultimately was created,

the adaptive neuro fuzzy

inference system design

was based on PI controllers cloning and arranged in an advance override

structure with the

aim of

keeping the optimum inlet-outlet pressures of the pump avoiding the

“Slug-Flow”

hydraulic phenomenon in the higher point of altimetric profile. For testing the hypothesis, the override neuro-fuzzy structure was always simulated considering the control by outlet pressure of the pump; nevertheless, when there were errors in pump outlet pressure setpoint takes the control by pump inlet pressure; in addiction to that, was able to soften the disturbance on the pump outlet up to -30psig considering them with a little instrumentation fault or hydraulic phenomenon, a greater reduction of pressure could be considered as a real ruptured pipeline; therefore the automatic control was lost . Finally, an investigation is recommended for the design of neuro-fuzzy controllers, taking as reference the critical points of the pipeline in order to mitigate their harmful phenomena.